Trapped in a Soundproof Box: How CO2 Levels Influence Speech Quality Assessment and Response Behaviour

To understand how we judge the quality of a phone call or a voice message, scientists often rely on a simple but rigorous process: they ask human listeners to sit in a quiet room and rate the sound. These listening tests are the gold standard for figuring out if a new technology makes speech sound clearer or more distorted. The assumption has always been that if the room is quiet and the sound is controlled, the listener's judgment reflects only the quality of the audio. However, this assumption overlooks a silent variable that lives inside every closed room: the air itself. As people breathe, they exhale carbon dioxide, a gas that accumulates in spaces with poor ventilation. While we often think of air quality in terms of dust or smells, scientists have found that rising levels of carbon dioxide can subtly change how our brains work, affecting our ability to make complex decisions and stay focused. If the air in a testing room becomes stale, it might not just make a person feel tired; it could actually alter the data they produce, turning a precise scientific measurement into something biased by the very environment it was meant to control.

A team of researchers at the Technische Universität Berlin decided to test this idea directly. They set up a listening experiment inside a soundproof booth, a small, sealed room designed to block out all outside noise. The goal was to evaluate the quality of various speech samples, ranging from crystal clear to heavily distorted. To do this, they invited twenty-four participants to sit inside the booth and listen to hundreds of short audio clips. The participants had to rate each clip on a scale, deciding how good or bad the speech sounded. Crucially, the researchers turned off the booth's ventilation system. This was not an accident; the ventilation fan was too noisy for the delicate listening task, so the team had to leave the door closed. As the session went on, the participants breathed in and out, and the concentration of carbon dioxide inside the small room began to climb naturally, rising from a fresh level of about 458 parts per million to over 1,500 parts per million within an hour. The researchers tracked this rise minute by minute, matching every single rating the participants gave to the exact level of carbon dioxide in the air at that moment.

What the researchers found was a clear pattern that challenges the idea of a neutral testing environment. As the air inside the booth became stale and the carbon dioxide levels rose, the way people rated the sound changed in a very specific way. First, the listeners began to make their decisions faster. They did not take as much time to think about whether a sound was good or bad; instead, they seemed to rush through the ratings, suggesting a shift toward more impulsive thinking. More importantly, their ratings became less precise. When the air was fresh, listeners could clearly tell the difference between excellent speech and poor speech. They used the full range of the rating scale, giving high scores to the best sounds and low scores to the worst. But as the carbon dioxide levels increased, this ability to distinguish quality faded. The listeners started to cluster their ratings toward the middle of the scale.

This shift created a strange distortion in the results. High-quality speech samples, which should have received top marks, were rated lower than they should have been. Conversely, poor-quality speech samples, which should have been rated very low, were given higher, more forgiving scores. The listeners seemed to lose their sensitivity to the extremes. Instead of saying "this is terrible" or "this is perfect," they began to say, "this is okay," regardless of what they were actually hearing. The researchers also noticed that the group of listeners became more uniform in their mistakes. When the air was fresh, different people had different opinions, leading to a wide spread of scores. As the air quality worsened, everyone's ratings moved closer together, becoming less varied and less diverse. It was as if the stale air smoothed out the edges of their perception, making everyone agree on a mediocre middle ground rather than holding firm to their individual judgments.

The study suggests that this effect is not just about feeling sleepy or uncomfortable. The researchers propose that the brain, when faced with the extra cognitive load of breathing in stale air, may be trying to save energy. Instead of carefully analyzing every nuance of the sound, the listener might default to a simpler, easier strategy: picking a middle number. This is a form of mental shortcut, where the brain avoids the effort of making a difficult, fine-grained distinction. The findings indicate that the environment in which a test takes place is not a passive backdrop but an active participant in the results. If a listening test is conducted in a room that is not properly ventilated, the data collected may be systematically biased, making good sounds look worse and bad sounds look better.

For the field of audio research, this discovery highlights a critical need for better experimental design. The researchers argue that future studies must treat air quality as a key variable, just like they treat background noise or the quality of the speakers. In long listening sessions, especially in small, sealed rooms, the air can become thick with carbon dioxide quickly, potentially skewing the data before the experiment is even finished. The solution is not necessarily to build new rooms, but to ensure that existing ones are ventilated properly, or to monitor the air levels in real time to account for any changes. The study concludes that ensuring high-quality experimental data requires more than just careful sound design and strict procedures; it also requires clean, breathable air. Without it, the very act of listening to a sound might be compromised by the air the listener is breathing.